A Genetic Algorithm and neural network hybrid classification scheme
David Decker, Joseph Hintz · 9th Computing in Aerospace Conference · 1993
This paper describes an approach to the classification of signals, based on the integration of Genetic Algorithms with Neural Network learning algorithms. First, a Genetic Algorithm is applied to the off-line task of selecting a minimal feature set, for later use in the classifier system. The second use of the Genetic Algorithm, in conjunction with the Backpropagation algorithm, is improvement in the training of multi-layer network weights. We demonstrate the effectiveness of this approach on a complex target classification task, and the promising result is generally applicable to a wide spectrum of signal classification tasks. By integrating the Genetic Algorithm with various Neural Network algorithms, we demonstrate a high degree of synergy which allows improved learning and performance, in both the extraction of key data features, and the classification of individuals based on these features.